Analysis of the Government Pension Fund of Norway
Yihong Zuo PB22000116
Yinan Bu PB22071444
1 Background
1.1 About the Norwegian Pension Fund
Norway’s pension system is globally recognized for its sustainability, transparency, and fiscal discipline. At its core lies the Government Pension Fund Global (GPFG)—commonly referred to as the Norwegian Oil Fund—which is the largest sovereign wealth fund (SWF) in the world, with assets exceeding $1.5 trillion USD.
Established in the 1990s to manage the country’s oil revenues, the fund operates under a long-term investment strategy designed to ensure that current and future generations benefit equally from Norway’s petroleum wealth. Managed by Norges Bank Investment Management (NBIM), the fund is invested globally across equities, fixed income, and real estate, and its returns form a key pillar of the Norwegian welfare state.
1.2 🧭 Research Motivation
Sovereign wealth funds represent a unique class of institutional investors: while they resemble mutual funds and pension funds in their diversified investment strategies, they also carry macroeconomic, geopolitical, and policy-making implications. Among them, Norway’s GPFG stands out as a model of fiscal prudence and ethical investment governance.
Our research is motivated by several key factors:
Growing global attention on SWFs: In an era marked by rising economic nationalism and geopolitical tensions—from the Trump-era tariff wars to ongoing resource competition—sovereign funds have become increasingly relevant strategic actors in global finance.
NBIM as a policy and performance benchmark: As the world’s largest SWF, GPFG sets global standards for transparency, ESG investment, and long-term capital stewardship.
Academic interest in bridging finance and public economics: Unlike private financial instruments, sovereign funds directly interface with national fiscal policies, intergenerational equity, and global political dynamics.
Through this lens, Norway’s pension model offers a rare opportunity to study the intersection of state power, financial strategy, and social contract.
2 Data Sources
2.1 Data Introduction
This study is based on official data from the Norwegian Government Pension Fund Global (GPFG), managed by Norges Bank Investment Management (NBIM). The fund aims to ensure long-term value creation for future generations in Norway by maintaining a globally diversified portfolio. The analysis uses annual investment records from 1998 to 2024, covering the fund’s major asset classes. Key data characteristics and current portfolio composition are outlined as follows:
- Equities(eq)
Equities represent ownership in more than 8,500 companies across global markets. As of the latest reporting period, they constitute 71.4% of the fund’s total investments, equivalent to approximately 14,113 billion kroner. Equity data include market values and sector-level exposures by country and year. This asset class provides the fund with long-term growth potential and broad economic participation.
- Fixed Income(fi)
This category consists primarily of government and corporate bonds, contributing 26.6% of the fund’s portfolio, or about 5,253 billion kroner. Fixed income assets offer income stability and diversification benefits, particularly during periods of equity market volatility. The data include issuer types, maturity profiles, and currency allocations.
- Real Estate(re)
The fund’s real estate investments focus on unlisted office buildings, logistics centers, and retail properties located in key international cities. This asset class accounts for 1.8% of total holdings, with a valuation of 364 billion kroner. Real estate exposure data include regional allocations, asset categories, and valuation metrics.
- Infrastructure(inf)
Infrastructure investments are concentrated in renewable energy projects, such as onshore wind and solar power. Although relatively new, this category now represents 0.1% of the portfolio, valued at 25 billion kroner. These investments align with the fund’s broader goals of sustainability and long-term resilience. The infrastructure dataset covers project types, energy sources, and investment geography.
All datasets used in this study are obtained from NBIM’s publicly available annual disclosures. Each file contains standardized, machine-readable tables covering country or issuer-level detail across market value, currency, and sector. These files can be batch downloaded via uniform resource links for consistent and replicable analysis.
2.2 Data Pre-processing
To prepare the data for analysis, we developed a general-purpose
preprocessing function (process_nbim_data)
which: - Downloads the raw Excel files for a specified asset class and
time range
- Extracts relevant fields such as market value, country, and year
- Aggregates the data to compute the total investment value per
country per year
- Outputs a clean and consistent CSV summary table
This function ensures comparability across asset classes and time periods, and enables further analysis such as cross-country comparisons and time series trends.
library(httr)
library(readxl)
library(readr)
library(dplyr)
library(tidyr)
process_nbim_data <- function(category,
currency = "NOK",
years = 2001:2024,
output_dir = "nbim_data",
group_by_level = "country") {
currency <- toupper(currency)
group_by_level <- tolower(group_by_level)
if (!currency %in% c("NOK", "USD")) {
stop("Currency must be either 'NOK' or 'USD'")
}
if (!group_by_level %in% c("country", "industry")) {
stop("group_by_level must be either 'country' or 'industry'")
}
dir.create(output_dir, showWarnings = FALSE)
data_list <- list()
for (year in years) {
cat("Processing", category, year, "...\n")
url <- paste0("https://www.nbim.no/hr/report/?category=", category, "&sortby=country&filetype=xlsx&year=", year)
destfile <- file.path(output_dir, paste0(category, "_", year, ".xlsx"))
tryCatch({
response <- GET(url, write_disk(destfile, overwrite = TRUE))
if (response$status_code != 200) {
cat("Failed to download year", year, "\n")
next
}
}, error = function(e) {
cat("Error downloading", year, "\n")
next
})
df <- tryCatch({
read_excel(destfile, sheet = excel_sheets(destfile)[1])
}, error = function(e) {
cat("Error reading Excel for", year, "\n")
return(NULL)
})
if (!is.null(df)) {
df$Year <- year
data_list[[as.character(year)]] <- df
}
}
combined <- bind_rows(data_list)
if (category %in% c("eq", "fi")) {
value_col <- ifelse(currency == "NOK", "Market Value(NOK)", "Market Value(USD)")
group_col <- ifelse(group_by_level == "industry", "Industry", "Country")
if (!value_col %in% names(combined)) {
stop(paste("Column", value_col, "not found in data"))
}
if (!group_col %in% names(combined)) {
stop(paste("Grouping column", group_col, "not found in data"))
}
combined <- combined %>%
mutate(
!!value_col := as.numeric(.data[[value_col]]),
Year = as.integer(Year)
) %>%
filter(!is.na(.data[[value_col]]), !is.na(.data[[group_col]]), !is.na(Year)) %>%
group_by(.data[[group_col]], Year) %>%
summarise(`Total Market Value` = sum(.data[[value_col]], na.rm = TRUE), .groups = "drop") %>%
rename(Group = 1) %>%
pivot_wider(names_from = Year, values_from = `Total Market Value`, names_sort = TRUE)
combined[is.na(combined)] <- 0
output_file <- file.path(output_dir, paste0(category, "_by_", group_by_level, "(", currency, ").csv"))
} else if (category %in% c("re", "inf")) {
value_col <- ifelse(currency == "NOK", "Total country value", "Total country value (USD)")
if (!value_col %in% names(combined)) {
stop(paste("Column", value_col, "not found in data"))
}
combined <- combined %>%
mutate(
!!value_col := as.numeric(gsub(",", "", .data[[value_col]])),
Year = as.integer(Year)
) %>%
filter(!is.na(.data[[value_col]]), !is.na(Country), !is.na(Year)) %>%
group_by(Country, Year) %>%
summarise(`Total Market Value` = sum(.data[[value_col]], na.rm = TRUE), .groups = "drop") %>%
rename(Group = Country) %>%
pivot_wider(names_from = Year, values_from = `Total Market Value`, names_sort = TRUE)
combined[is.na(combined)] <- 0
output_file <- file.path(output_dir, paste0(category, "_by_country(", currency, ").csv"))
} else {
stop("Unsupported category: must be one of 'eq', 'fi', 're', 'inf'")
}
write.csv(combined, output_file, row.names = FALSE)
cat("✅ Finished processing", category, "in", currency, "grouped by", group_by_level, "and saved to", output_file, "\n")
return(combined)
}Then we can easily get the summary imformation by using this function. The following is a sample to get the equities ivestment summary.
equities_summary <- process_nbim_data(
category = "fi", #
years = 1998:2024,
output_dir = "nbim_data",
currency = "USD",
group_by_level = "Industry"
)- Statistics Norway (SSB)
- OECD Pension Database
- Norwegian Labour and Welfare Administration (NAV)
- Other academic and policy sources
3 Equity Investment Analysis
3.1 Equity Investment by Country
3.1.1 Country Share Over Time
The stacked area chart below illustrates the Norwegian Pension Fund’s annual equity investment share across its top invested countries, including China. By visualizing the capital allocation from year to year, we can observe clear trends such as increasing concentration in specific countries and shifts in geographic investment priorities. This provides a comprehensive view of how the fund’s international exposure has evolved over time.
library(tidyverse)
library(scales)
library(readr)
equities <- read_csv("nbim_data/eq_by_country(USD).csv")equities_long <- equities %>%
pivot_longer(-Country, names_to = "Year", values_to = "Investment") %>%
mutate(Year = as.integer(Year)) %>%
filter(!is.na(Investment))
equities_long$Investment[is.na(equities_long$Investment)] <- 0
total_investment_per_year <- equities_long %>%
group_by(Year) %>%
summarise(Total = sum(Investment))
equities_share <- equities_long %>%
left_join(total_investment_per_year, by = "Year") %>%
mutate(Share = Investment / Total)
# 1️⃣ Visualization: Yearly Investment Share for Top Countries + China
top_countries <- equities_long %>%
group_by(Country) %>%
summarise(TotalInvestment = sum(Investment)) %>%
top_n(6, TotalInvestment) %>%
pull(Country)
# Ensure China is included
top_countries <- union(top_countries, "China")
library(RColorBrewer)
pastel_colors <- brewer.pal(n = 8, name = "Set2")
ggplot(
equities_share %>% filter(Country %in% top_countries),
aes(x = Year, y = Share, fill = Country)
) +
geom_area(alpha = 0.9, color = "white", size = 0.3) +
scale_fill_manual(values = pastel_colors) +
scale_y_continuous(labels = percent_format()) +
labs(
title = "Norwegian Pension Fund Equity Investment – Country Share Over Time",
y = "Investment Share",
x = "Year"
) +
theme_minimal(base_family = "Arial") +
theme(
plot.title = element_text(face = "bold", size = 16, color = "#333333"),
axis.title = element_text(size = 12, color = "#444444"),
axis.text = element_text(color = "#666666"),
legend.position = "bottom",
legend.title = element_blank(),
panel.grid.major.x = element_blank(),
panel.grid.minor = element_blank()
)
3.1.2 Investment Concentration: Herfindahl-Hirschman Index (HHI)
The Herfindahl-Hirschman Index (HHI) is a well-established measure of portfolio concentration. It is computed as the sum of the squares of each country’s investment share in a given year: \[ \mathrm{HHI} = \sum_{i=1}^{n} s_i^2 \] where $ s_i $ is the proportion of the total investment allocated to country $ i $ in a specific year. An HHI value closer to 1 indicates a highly concentrated investment portfolio (i.e., focused in a few countries), while a value closer to 0 indicates a more diversified allocation.
hhi_per_year <- equities_share %>%
group_by(Year) %>%
summarise(HHI = sum(Share^2))
library(plotly)
library(RColorBrewer)
main_color <- brewer.pal(8, "Set2")[1]
p <- ggplot(hhi_per_year, aes(x = Year, y = HHI)) +
geom_line(color = main_color, size = 1) +
geom_point(color = main_color) +
labs(
title = "Yearly Investment Concentration (Herfindahl Index)",
y = "Herfindahl-Hirschman Index (HHI)",
x = "Year"
) +
theme_minimal()
ggplotly(p)3.1.3 Key Insights
- Rising Investment Concentration
The HHI chart reveals a notable increase in portfolio concentration starting around 2013, accelerating post-2020. This suggests a strategic shift by the fund towards fewer, larger markets — a possible response to geopolitical tensions, market volatility, or a broader trend of deglobalization. Notably, the U.S. has become an increasingly dominant destination.
- China as a Strategic Outlier
Despite the general move toward concentration, China’s share in the fund has steadily increased. Unlike other emerging markets whose shares have stagnated or declined, China’s growing role in the portfolio underscores its strategic significance. This may reflect confidence in China’s long-term growth prospects and its centrality in global supply chains, even amid rising geopolitical risks.
3.2 Equity Investment by Industry
3.2.1 Data Preparation and Consumer Sector Consolidation
To provide clearer insights, all industry groups containing “consumer” (regardless of case) were aggregated into a single “Total Consumer” category. This ensures a more coherent and interpretable representation of sectoral investment trends.
library(tidyverse)
library(scales)
library(readr)
equities <- read_csv("nbim_data/eq_by_industry(USD).csv")library(dplyr)
library(stringr)
library(dplyr)
library(stringr)
consumer_rows <- equities %>%
filter(str_detect(Group, regex("consumer", ignore_case = TRUE)))
consumer_total <- consumer_rows %>%
select(-Group) %>%
summarise(across(everything(), sum, na.rm = TRUE))
total_consumer_row <- cbind(Group = "Total Consumer", consumer_total)
equities_cleaned <- equities %>%
filter(!str_detect(Group, regex("consumer", ignore_case = TRUE)))
equities_final <- bind_rows(equities_cleaned, total_consumer_row)
equities <- equities_final[, names(equities)]equities_long <- equities %>%
pivot_longer(-Group, names_to = "Year", values_to = "Investment") %>%
mutate(Year = as.integer(Year)) %>%
filter(!is.na(Investment))
equities_long$Investment[is.na(equities_long$Investment)] <- 03.2.2 Industry Share Over Time
The following chart displays the fund’s equity allocation by industry over time. It highlights the six most heavily invested industries (post-aggregation), offering a clear perspective on how sectoral priorities have shifted.
total_investment_per_year <- equities_long %>%
group_by(Year) %>%
summarise(Total = sum(Investment))
equities_share <- equities_long %>%
left_join(total_investment_per_year, by = "Year") %>%
mutate(Share = Investment / Total)
# 1️⃣ Visualization: Yearly Investment Share for Top Countries + China
top_countries <- equities_long %>%
group_by(Group) %>%
summarise(TotalInvestment = sum(Investment)) %>%
top_n(6, TotalInvestment) %>%
pull(Group)
# Plot area chart
palette <- brewer.pal(min(length(top_countries), 8), "Set2")
if (length(top_countries) > 8) {
palette <- colorRampPalette(brewer.pal(8, "Set2"))(length(top_countries))
}
ggplot(
equities_share %>% filter(Group %in% top_countries),
aes(x = Year, y = Share, fill = factor(Group, levels = top_countries))
) +
geom_area(alpha = 0.8, color = "white", size = 0.3) +
scale_fill_manual(values = palette, name = NULL) +
scale_y_continuous(labels = percent_format()) +
labs(
title = "Norwegian Pension Fund Equity Investment – Country Share Over Time",
y = "Investment Share",
x = "Year"
) +
theme_minimal(base_family = "Arial") +
theme(
plot.title = element_text(face = "bold", size = 16, color = "#333333"),
axis.title = element_text(size = 12, color = "#444444"),
axis.text = element_text(color = "#666666"),
legend.position = "bottom",
panel.grid.major.x = element_blank(),
panel.grid.minor = element_blank()
)
3.2.3 Industry Concentration: Herfindahl Index
The sector-level HHI plot reflects the evolution of the fund’s diversification across industries.
hhi_per_year <- equities_share %>%
group_by(Year) %>%
summarise(HHI = sum(Share^2))
events <- data.frame(
start = c(2007, 2020),
end = c(2009, 2022),
label = c("Global Financial Crisis", "COVID-19 Pandemic")
)
ymin <- min(hhi_per_year$HHI, na.rm = TRUE)
ymax <- max(hhi_per_year$HHI, na.rm = TRUE)
p <- ggplot(hhi_per_year, aes(x = Year, y = HHI)) +
geom_rect(
data = events,
aes(xmin = start, xmax = end, ymin = ymin, ymax = ymax),
inherit.aes = FALSE,
fill = "gray80",
alpha = 0.4
) +
geom_line(color = main_color, size = 1) +
geom_point(color = main_color) +
geom_text(
data = events,
aes(x = (start + end) / 2, y = ymax, label = label),
inherit.aes = FALSE,
vjust = -0.5,
size = 3.5,
fontface = "italic"
) +
labs(
title = "Yearly Investment Concentration (Herfindahl Index)",
subtitle = "Shaded regions indicate major global events",
y = "Herfindahl-Hirschman Index (HHI)",
x = "Year"
) +
theme_minimal()
ggplotly(p)3.2.4 🔍 Key Insights: Industry Allocation of the Norwegian Pension Fund
- Stable Long-Term Allocation with Dominant Core Sectors
The stacked area chart shows that the fund’s industry allocation has remained relatively stable over time. Financials and Total Consumer (which aggregates consumer-related sectors) have consistently accounted for the largest shares, together making up over half of the portfolio in most years. Industrials and Technology have also grown in importance, particularly in the past decade.
✅ Balance Between Stability and Growth: The fund maintains strong exposure to traditional sectors like financials while gradually increasing investments in growth-oriented industries such as technology and healthcare.
2.Cyclical Fluctuations in Investment Concentration
The Herfindahl-Hirschman Index (HHI) trend reveals noticeable fluctuations in industry concentration from 2000 to 2025. Key observations:
2007–2009: A sharp drop in HHI during the global financial crisis, suggesting a diversification move to mitigate risk.
2013–2019: A steady rise in HHI indicates growing concentration in a few preferred sectors.
2020–2022: A dramatic dip during the COVID-19 pandemic, followed by a swift recovery in concentration.
🔄 Dynamic Strategy: The fund adjusts its level of industry concentration in response to global economic shocks, moving between diversification and focus depending on market conditions.
- Rising Importance of Technology and Healthcare
Although still smaller in absolute terms, Technology and Health Care sectors have seen a consistent increase in allocation over the past decade. This signals a strategic pivot toward long-term growth sectors, aligned with global trends in innovation and demographic shifts.
🚀 Forward-Looking Positioning: The fund is gradually expanding exposure to future-facing sectors, which may offer sustainable returns over the long term.
- Enhanced Clarity Through Consumer Sector Consolidation
By merging “Consumer Discretionary” and “Consumer Staples” into a unified Total Consumer category, the analysis presents a clearer picture of the sector’s true weight. The data confirms that consumer-related industries remain one of the most stable and substantial components of the portfolio.
4 Fixed Income Investment Analysis
4.1 Fixed Income Investment by Country
library(tidyverse)
library(scales)
library(readr)
fixed_income <- read_csv("nbim_data/fi_by_country(USD).csv")
names(fixed_income)[1] <- "Country"
fi_long <- fixed_income %>%
pivot_longer(-Country, names_to = "Year", values_to = "Investment") %>%
mutate(Year = as.integer(Year)) %>%
filter(!is.na(Investment))
fi_long$Investment[is.na(fi_long$Investment)] <- 0
total_investment_per_year <- fi_long %>%
group_by(Year) %>%
summarise(Total = sum(Investment))
fi_share <- fi_long %>%
left_join(total_investment_per_year, by = "Year") %>%
mutate(Share = Investment / Total)top_countries <- fi_long %>%
group_by(Country) %>%
summarise(TotalInvestment = sum(Investment)) %>%
top_n(6, TotalInvestment) %>%
pull(Country)
top_countries <- union(top_countries, "China")
palette <- brewer.pal(min(length(top_countries), 8), "Set2")
if (length(top_countries) > 8) {
palette <- colorRampPalette(brewer.pal(8, "Set2"))(length(top_countries))
}
ggplot(
fi_share %>% filter(Country %in% top_countries),
aes(x = Year, y = Share, fill = Country)
) +
scale_fill_manual(values = palette, name = NULL) +
geom_area(alpha = 0.8) +
scale_y_continuous(labels = percent_format()) +
labs(
title = "Norwegian Pension Fund – Fixed Income Investment by Country",
y = "Investment Share",
x = "Year"
) +
theme_minimal()
hhi_per_year <- fi_share %>%
group_by(Year) %>%
summarise(HHI = sum(Share^2))
events <- data.frame(
start = c(2007, 2020),
end = c(2009, 2022),
label = c("Global Financial Crisis", "COVID-19 Pandemic")
)
main_color <- brewer.pal(8, "Set2")[2]
ymin <- min(hhi_per_year$HHI, na.rm = TRUE)
ymax <- max(hhi_per_year$HHI, na.rm = TRUE)
events$ymin <- ymin
events$ymax <- ymax
p <- ggplot(hhi_per_year, aes(x = Year, y = HHI)) +
geom_rect(
data = events,
aes(xmin = start, xmax = end, ymin = ymin, ymax = ymax),
inherit.aes = FALSE,
fill = "gray80",
alpha = 0.4
) +
geom_line(color = main_color, size = 1) +
geom_point(color = main_color) +
geom_text(
data = events,
aes(x = (start + end)/2, y = ymax, label = label),
inherit.aes = FALSE,
vjust = -0.5,
size = 3.5,
fontface = "italic"
) +
labs(
title = "Fixed Income Investment Concentration (Herfindahl Index)",
subtitle = "Shaded areas mark major global events",
y = "Herfindahl-Hirschman Index (HHI)",
x = "Year"
) +
theme_minimal()
ggplotly(p)4.1.1 Insights
Diversification: The chart shows that the Norwegian Pension Fund’s fixed income investments are spread across multiple countries, indicating a diversified portfolio.
Temporal Changes: While the fund’s allocation fluctuates over time, the share of some countries appears to increase or decrease. For example, the United States (pink) consistently occupies a large portion of the share, especially after 2008.
Post-2008 Patterns: After the Global Financial Crisis, some changes in allocation might be noticeable, suggesting a potential reallocation strategy in response to market conditions.
Recent Increase: Around 2020 onwards, the overall investment shares remain relatively stable with moderate fluctuations, though slight changes might be observed during the COVID-19 period.
4.2 Fixed Income Investment by Industy
fixed_income <- read_csv("nbim_data/fi_by_industry(USD).csv")
consumer_rows <- fixed_income %>%
filter(str_detect(Group, regex("consumer", ignore_case = TRUE)))
consumer_total <- consumer_rows %>%
select(-Group) %>%
summarise(across(everything(), sum, na.rm = TRUE))
total_consumer_row <- cbind(Group = "Total Consumer", consumer_total)
fixed_income_cleaned <- fixed_income %>%
filter(!str_detect(Group, regex("consumer", ignore_case = TRUE)))
fixed_income_final <- bind_rows(fixed_income_cleaned, total_consumer_row)
fixed_income <- fixed_income_final[, names(fixed_income)]
fi_long <- fixed_income %>%
pivot_longer(-Group, names_to = "Year", values_to = "Investment") %>%
mutate(Year = as.integer(Year)) %>%
filter(!is.na(Investment))
fi_long$Investment[is.na(fi_long$Investment)] <- 0
total_investment_per_year <- fi_long %>%
group_by(Year) %>%
summarise(Total = sum(Investment))
fi_share <- fi_long %>%
left_join(total_investment_per_year, by = "Year") %>%
mutate(Share = Investment / Total)
top_groups <- fi_long %>%
group_by(Group) %>%
summarise(TotalInvestment = sum(Investment)) %>%
top_n(6, TotalInvestment) %>%
pull(Group)
library(ggplot2)
library(RColorBrewer)
library(scales)
# 自动适应分组数的配色方案
palette <- brewer.pal(min(length(top_groups), 8), "Set2")
if (length(top_groups) > 8) {
palette <- colorRampPalette(brewer.pal(8, "Set2"))(length(top_groups))
}
ggplot(
fi_share %>% filter(Group %in% top_groups),
aes(x = Year, y = Share, fill = factor(Group, levels = top_groups))
) +
geom_area(alpha = 0.8, color = "white", size = 0.3) +
scale_fill_manual(values = palette, name = NULL) +
scale_y_continuous(labels = percent_format()) +
labs(
title = "Norwegian Pension Fund – Fixed Income by Sector (Top 6 incl. Consumer)",
y = "Investment Share",
x = "Year"
) +
theme_minimal(base_family = "Arial") +
theme(
plot.title = element_text(face = "bold", size = 16, color = "#333333"),
axis.title = element_text(size = 12, color = "#444444"),
axis.text = element_text(color = "#666666"),
legend.position = "bottom",
panel.grid.major.x = element_blank(),
panel.grid.minor = element_blank()
)
hhi_per_year <- fi_share %>%
group_by(Year) %>%
summarise(HHI = sum(Share^2))
main_color <- brewer.pal(8, "Set2")[3]
ymin <- min(hhi_per_year$HHI, na.rm = TRUE)
ymax <- max(hhi_per_year$HHI, na.rm = TRUE)
events$ymin <- ymin
events$ymax <- ymax
p <- ggplot(hhi_per_year, aes(x = Year, y = HHI)) +
geom_rect(
data = events,
aes(xmin = start, xmax = end, ymin = ymin, ymax = ymax),
inherit.aes = FALSE,
fill = "gray80",
alpha = 0.4
) +
geom_line(color = main_color, size = 1) +
geom_point(color = main_color) +
geom_text(
data = events,
aes(x = (start + end)/2, y = ymax, label = label),
inherit.aes = FALSE,
vjust = -0.5,
size = 3.5,
fontface = "italic"
) +
labs(
title = "Fixed Income Sector Concentration (Herfindahl Index)",
subtitle = "With crisis periods highlighted",
y = "HHI",
x = "Year"
) +
theme_minimal()
ggplotly(p)4.2.1 Insights
1995–2000: Very high HHI (~0.9) → Highly concentrated investments.
2000–2005: Sharp decline in HHI → Increased diversification.
2005–2015: Relatively stable, moderate HHI → Consistent diversification.
2008–2009 (Global Financial Crisis) and 2020–2022 (COVID-19 Pandemic): Small changes in HHI, indicating resilience in diversification strategy.
2023–2025: Slight uptick in HHI → Possible mild reconsolidation or strategic shifts.
Summary: The Norwegian Pension Fund moved from a concentrated to a more diversified fixed income portfolio over time, with only moderate concentration increases during recent years.
5 Return Analysis
5.1 Overview
This section provides a comprehensive analysis of the Norwegian Government Pension Fund Global (NBIM)’s annual return performance. As one of the largest sovereign wealth funds globally, NBIM’s long-term return pattern is critical to assessing the effectiveness, resilience, and global alignment of its investment strategy. By visualizing its historical annual returns, we gain an initial understanding of how the fund navigates both market expansions and downturns.
library(readxl)
library(quantmod)
library(dplyr)
# 1. Load NBIM return data
nbim_file <- "nbim_data/nbim-1749105481449.xls"
nbim_data <- read_excel(nbim_file)
colnames(nbim_data) <- c("Year", "NBIM_Return", "Accumulated")
# Ensure numeric and clean
nbim_data <- nbim_data %>%
filter(!is.na(Year)) %>%
mutate(NBIM_Return = NBIM_Return / 100)
p <- ggplot(nbim_data, aes(x = Year, y = NBIM_Return)) +
geom_col(fill = "steelblue") +
geom_hline(yintercept = 0, color = "black", linetype = "dashed") +
scale_y_continuous(labels = percent_format(accuracy = 1)) +
labs(
title = "Norwegian Pension Fund Annual Returns",
x = "Year",
y = "Annual Return"
) +
theme_minimal()
ggplotly(p)5.2 Robust Investment Strategy: Global Benchmark Comparison
5.2.1 🔍 Motivation
To further evaluate the robustness and global alignment of NBIM’s investment strategy, we compare its annual return with that of the S&P 500 Total Return Index from 1998 to 2024. The S&P 500 serves here as a proxy for global equity market performance, given its high correlation with the MSCI ACWI and its role as a widely tracked benchmark reflecting large-cap equity trends.
The central question we aim to address is: Does NBIM perform in line with global equity markets, and how strong is that relationship?
📊 Data Sources and Interpretation
NBIM Return Data: Collected from official NBIM reports, converted to annual percentage change format.
S&P 500 Total Return Data: Annual returns (including reinvested dividends) from 1998 to 2024, compiled from ETF-based total return proxies.
By aligning both datasets year by year, we construct a time series that allows for a direct comparison between NBIM and global equity performance.
5.2.2 🔍 Visual Comparison
As shown in the chart above, the return series of NBIM and the S&P 500 exhibit strikingly similar movements. Periods of sharp drawdowns (e.g., 2008, 2022) and market rallies (e.g., 2003, 2019, 2023) are strongly aligned, reflecting the fund’s high global market exposure, especially to large-cap equities.
# 加载必要包
library(readxl)
library(dplyr)
library(ggplot2)
library(scales)
nbim_file <- "nbim_data/nbim-1749105481449.xls"
nbim_data <- read_excel(nbim_file)
colnames(nbim_data) <- c("Year", "NBIM_Return", "Accumulated")
nbim_data <- nbim_data %>%
filter(!is.na(Year)) %>%
mutate(
Year = as.numeric(Year),
NBIM_Return = NBIM_Return / 100
)
sp500_file <- "nbim_data/sp500_returns_1998_2024.csv"
sp500_data <- read.csv(sp500_file)
comparison_df <- inner_join(nbim_data, sp500_data, by = "Year")
comparison_long <- comparison_df %>%
select(Year, NBIM_Return, SP500_Return) %>%
tidyr::pivot_longer(cols = c("NBIM_Return", "SP500_Return"),
names_to = "Series",
values_to = "Return")
p <- ggplot(comparison_long, aes(x = Year, y = Return, color = Series)) +
geom_line(size = 1.2) +
geom_point() +
scale_y_continuous(labels = percent_format(accuracy = 1)) +
labs(
title = "Annual Returns: NBIM vs S&P 500 (1998–2024)",
x = "Year",
y = "Return",
color = "Series"
) +
theme_minimal()
ggplotly(p)5.2.3 📊 Statistical Summary
Pearson Correlation Coefficient: 0.918
This indicates a very strong positive linear relationship between the returns of NBIM and the S&P 500. In other words, when the S&P 500 performs well, NBIM tends to perform well too — and vice versa.
correlation <- cor(comparison_df$NBIM_Return, comparison_df$SP500_Return)
cat("📈 Correlation between NBIM and S&P 500 returns:", round(correlation, 3), "\n")## 📈 Correlation between NBIM and S&P 500 returns: 0.918cor.test(comparison_df$NBIM_Return, comparison_df$SP500_Return)##
## Pearson's product-moment correlation
##
## data: comparison_df$NBIM_Return and comparison_df$SP500_Return
## t = 11.614, df = 25, p-value = 1.443e-11
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.8272122 0.9625424
## sample estimates:
## cor
## 0.9184952The correlation analysis between the Norwegian Government Pension Fund Global (NBIM) and the S&P 500 reveals a strong and statistically significant relationship, with a Pearson correlation coefficient of 0.918 (95% CI: 0.827 to 0.963, p < 0.001). This confirms that NBIM’s annual equity returns closely track global equity markets, particularly large-cap performance trends represented by the S&P 500.
Such a high correlation—combined with NBIM’s consistently competitive performance—highlights the fund’s robust and disciplined investment strategy. Rather than chasing short-term gains or relying on speculative positions, NBIM maintains a strategic, globally diversified portfolio that mirrors long-term market trends while incorporating prudent risk management. This approach not only provides stable returns over time but also enhances the fund’s public legitimacy and trust, as it demonstrates transparency, professionalism, and alignment with global economic fundamentals.
In times of market stress, NBIM often exhibits slightly lower volatility compared to the S&P 500, underscoring its resilience and the strength of its asset allocation framework. This consistency and discipline enable the fund to fulfill its long-term obligations for future generations, while maintaining public confidence and a positive reputation on the global financial stage.
5.3 Excess Return and Alpha Analysis
5.3.1 Excess Return: Annual Difference
To evaluate NBIM’s performance beyond simply tracking global equity markets, we calculate its annual excess return over the S&P 500. Excess return is defined as the annual return of NBIM minus the return of the S&P 500 for each year. A positive excess return indicates outperformance relative to the global benchmark.
comparison_df <- comparison_df %>%
mutate(ExcessReturn = NBIM_Return - SP500_Return)
p=ggplot(comparison_df, aes(x = Year, y = ExcessReturn, fill = ExcessReturn > 0)) +
geom_col(show.legend = FALSE) +
scale_fill_manual(values = c("TRUE" = "forestgreen", "FALSE" = "indianred")) +
scale_y_continuous(labels = percent_format(accuracy = 1)) +
labs(
title = "NBIM Excess Return vs S&P 500 (1998–2024)",
x = "Year",
y = "Excess Return (NBIM - S&P 500)"
) +
theme_minimal()
ggplotly(p)5.3.2 Alpha and Beta: Linear Regression
To further quantify whether NBIM is simply tracking the S&P 500 or providing additional value, we run a linear regression of NBIM annual returns against S&P 500 returns:
reg <- lm(NBIM_Return ~ SP500_Return, data = comparison_df)
summary(reg)##
## Call:
## lm(formula = NBIM_Return ~ SP500_Return, data = comparison_df)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.072794 -0.023620 -0.000704 0.018571 0.102134
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.012678 0.009475 1.338 0.193
## SP500_Return 0.534345 0.046009 11.614 1.44e-11 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.04235 on 25 degrees of freedom
## Multiple R-squared: 0.8436, Adjusted R-squared: 0.8374
## F-statistic: 134.9 on 1 and 25 DF, p-value: 1.443e-11Insignts The linear regression of NBIM’s annual returns on the S&P 500 Total Return Index for 1998–2024 yields the following results:
Intercept (Alpha): 0.0127 (or 1.27%)
Beta: 0.534
R-squared: 0.844
p-value for S&P 500 coefficient: < 0.001
Interpretation:
Alpha (Intercept): The estimated intercept of 1.27% suggests that, on average, NBIM generated an annual return that is approximately 1.27 percentage points higher than what would be predicted solely by S&P 500 movements. However, this alpha is not statistically significant (p = 0.193), so we cannot conclusively say NBIM consistently outperformed the benchmark after adjusting for market exposure.
Beta: The beta of 0.53 indicates that NBIM’s returns are much less sensitive to the S&P 500’s fluctuations. A beta below 1 implies that NBIM has significantly lower volatility than the global equity market, consistent with its globally diversified and multi-asset portfolio. This defensive characteristic means NBIM typically participates less in extreme market ups and downs.
R-squared: The R-squared of 0.844 shows that approximately 84% of the variability in NBIM’s returns can be explained by movements in the S&P 500. This high value reflects a strong connection to global equity trends, though not a full replication of the index.
Significance: The coefficient for S&P 500 returns is highly significant (p < 0.001), indicating a robust statistical relationship between NBIM and global equity market performance.
5.3.3 Conclusion:
Overall, NBIM’s returns are strongly, but not perfectly, linked to the global equity market, with substantially lower sensitivity (beta) and volatility than the S&P 500. While the estimated alpha is positive, it is not statistically significant, so NBIM’s main value-add appears to be in risk reduction and diversification, rather than in delivering consistent excess returns above the benchmark.
6 Industry Allocation Clustering Analysis: Based on k-Means
While sector-level investment concentration can be measured using indices like the Herfindahl-Hirschman Index (HHI), such metrics do not reveal how similar or different the fund’s industry allocation structure has been across years. To uncover patterns in the fund’s strategic orientation toward different sectors over time, we perform k-means clustering on the yearly equity allocations by industry.
Why This Matters
Helps identify structurally similar periods in terms of sector exposure;
Detects shifts in strategic focus, such as increasing emphasis on technology or de-risking from cyclical industries;
Reveals how the fund’s sectoral priorities evolve, especially during market disruptions like the 2008 crisis or the COVID-19 pandemic.
6.1 Methodology
Data Preparation: We calculate the relative share of investment in each industry for every year, using cleaned and aggregated data (including a consolidated “Total Consumer” sector). This results in a matrix where each row corresponds to a year, and each column represents an industry’s share.
Clustering Approach: We apply k-means clustering to group years based on the similarity of their industry allocation profiles. The optimal number of clusters is selected using the elbow method, balancing interpretability and within-cluster consistency.
Interpretation: Each cluster captures a unique sectoral allocation profile that reflects strategic themes — such as “broadly diversified,” “tech-heavy,” or “financials-dominant.” Comparing these clusters with historical events and concentration trends provides deeper insight into the fund’s evolving investment logic.
By clustering years based on their sector allocation patterns, we move beyond single metrics and gain a multidimensional view of investment strategy across time.
library(tidyverse)
library(cluster)
library(factoextra)
library(scales)
equities <- read_csv("nbim_data/eq_by_industry(USD).csv")
consumer_rows <- equities %>%
filter(str_detect(Group, regex("consumer", ignore_case = TRUE)))
consumer_total <- consumer_rows %>%
select(-Group) %>%
summarise(across(everything(), sum, na.rm = TRUE))
total_consumer_row <- cbind(Group = "Total Consumer", consumer_total)
equities_cleaned <- equities %>%
filter(!str_detect(Group, regex("consumer", ignore_case = TRUE)))
equities_final <- bind_rows(equities_cleaned, total_consumer_row)
equities <- equities_final[, names(equities_final)] # Ensure correct order
equities_long <- equities %>%
pivot_longer(-Group, names_to = "Year", values_to = "Investment") %>%
mutate(Year = as.integer(Year)) %>%
filter(!is.na(Investment))
total_investment_per_year <- equities_long %>%
group_by(Year) %>%
summarise(Total = sum(Investment))
equities_share <- equities_long %>%
left_join(total_investment_per_year, by = "Year") %>%
mutate(Share = Investment / Total)
industry_matrix <- equities_share %>%
select(Year, Group, Share) %>%
pivot_wider(names_from = Group, values_from = Share, values_fill = 0)
industry_data <- industry_matrix %>% select(-Year) %>% as.matrix()
fviz_nbclust(industry_data, kmeans, method = "wss") +
labs(title = "Elbow Method: Optimal Number of Clusters (Industry)")
set.seed(42)
k_result <- kmeans(industry_data, centers = 4)
industry_matrix$Cluster <- factor(k_result$cluster)
p <- ggplot(industry_matrix, aes(x = Year, y = Cluster, color = as.factor(Cluster))) +
geom_point(size = 3, alpha = 0.85) +
labs(
title = "Year Clusters Based on Industry Allocation (Equity Investment)",
subtitle = "Each point represents a year; clusters reflect sector similarity",
x = "Year", y = "Cluster Group", color = "Cluster"
) +
theme_minimal(base_family = "Arial") +
theme(
plot.title = element_text(face = "bold", size = 16),
plot.subtitle = element_text(size = 12),
axis.text.y = element_text(face = "bold"),
legend.position = "right"
)
ggplotly(p)6.2 Analysis:
equities_with_cluster <- equities_share %>%
left_join(industry_matrix %>% select(Year, Cluster), by = "Year")cluster_labels <- tibble(
Cluster = factor(c(1, 2, 3, 4)),
ClusterLabel = c(
"2014–2020",
"1998–2006",
"2021–2024",
"2007–2013"
)
)
cluster_avg_share <- equities_with_cluster %>%
group_by(Cluster, Group) %>%
summarise(AvgShare = mean(Share, na.rm = TRUE), .groups = "drop") %>%
left_join(cluster_labels, by = "Cluster") # 加上带年份的标签
cluster_avg_share <- cluster_avg_share %>%
group_by(ClusterLabel) %>%
mutate(
PercentLabel = percent(AvgShare, accuracy = 1),
ypos = cumsum(AvgShare) - 0.5 * AvgShare
)
palette <- brewer.pal(min(length(unique(cluster_avg_share$Group)), 8), "Set2")
if (length(unique(cluster_avg_share$Group)) > 8) {
palette <- colorRampPalette(brewer.pal(8, "Set2"))(length(unique(cluster_avg_share$Group)))
}
clusters <- unique(cluster_avg_share$ClusterLabel)
plots <- list()
for (i in seq_along(clusters)) {
this_cluster <- clusters[i]
this_data <- cluster_avg_share %>% filter(ClusterLabel == this_cluster)
plots[[i]] <- plot_ly(
this_data,
labels = ~Group,
values = ~AvgShare,
type = 'pie',
textinfo = 'label+percent',
hoverinfo = 'label+percent',
marker = list(colors = palette),
showlegend = TRUE
) %>%
layout(
title = paste("Period:", this_cluster),
legend = list(orientation = "h", x = 0.1, y = -0.1)
)
}
plots[[2]]plots[[4]]plots[[1]]plots[[3]]📊 Chart Overview The pie charts display the average industry allocation within each cluster group derived from k-means clustering. Each cluster reflects a period with structurally similar sector exposures, revealing shifts in investment philosophy across time.
- 1998–2006: Traditional Diversification Key Characteristics:
High exposure to Financials and Industrials.
Total Consumer sector is also a major allocation.
Technology and Healthcare remain marginal.
Interpretation: This phase reflects a conventional asset allocation strategy, centered around cyclical and defensive industries. The portfolio likely aimed to balance risk and return through a broad sector spread, characteristic of early-stage fund structures.
- 2007–2013: Post-Crisis Rebalancing Key Characteristics:
Financials remained dominant, but the weight of Healthcare and Technology increased.
Still strong presence of Energy, Industrials, and Consumer sectors.
Interpretation: Following the 2008 Global Financial Crisis, the fund began reallocating toward resilient and growth-oriented sectors, such as Healthcare and Tech, while maintaining core exposure in traditional sectors. The structure suggests an emerging defensive tilt and diversification emphasis.
- 2014–2020: Technology Ascendance Key Characteristics:
Significant rise in Technology allocation — now second only to Financials.
Decline in Energy, Materials, and other cyclical sectors.
Consumer remains important, along with moderate Healthcare allocation.
Interpretation: This period marks a structural transition to tech-driven investing, aligned with the rise of FAANG stocks and the digital economy. The allocation suggests a deliberate overweight on high-growth, innovation-led industries.
- 2021–2024: Resilience & Sustainability Era Key Characteristics:
Strong allocations to Technology and Healthcare.
Lower weight in Energy and Industrials.
Total Consumer still features prominently.
Interpretation: Post-COVID, the portfolio reflects a resilience-first and sustainability-aware strategy. Increased weights in Healthcare and Tech point to a focus on structural stability, digital transformation, and possibly ESG-aligned sectors. The decline in traditional heavy industries signals a more forward-looking risk posture.
6.3 Key Takeaways
- Clear Evolution in Sector Strategy:
The portfolio transitioned from traditional finance-industrial dominance to tech-healthcare concentration.
- Crisis-Driven Allocation Shifts:
Both the 2008 and 2020 crises triggered reallocation toward defensive and structurally resilient sectors.
- Growing Emphasis on Growth and ESG Themes:
Technology has emerged as a strategic pillar, and later years show potential ESG alignment with reduced exposure to fossil-intensive sectors.
7 Shiny App
The Website is available at nbim.
7.1 Background
The management of money has been a subject of great interest and discussion throughout history. In today’s rapidly evolving global trade environment, the question of how to preserve and grow one’s wealth has become more relevant than ever. For individuals or institutions with assets, making informed decisions on how to allocate and invest resources wisely is both a fascinating and crucial challenge. The ultimate goal is to maintain asset value while maximizing returns through strategic investment.
7.1.1 Norwegian Government Pension Fund Global
The Norwegian Government Pension Fund Global is the world’s largest sovereign wealth fund, with assets exceeding 1.5 trillion USD. It represents the accumulated wealth of Norway’s fossil fuel revenues and serves as a long-term financial reserve for the country. The fund is owned by the Norwegian people, held in trust by the government, and managed by the Central Bank of Norway.
Each year, the Norwegian government withdraws approximately 3% of the fund’s value to support public spending, primarily for the benefit of its citizens through social welfare programs. This prudent and sustainable approach to wealth management is one of the key reasons why Norway, despite having a population of just 5.5 million, is renowned for offering some of the most comprehensive and generous social welfare systems in the world.
7.1.2 Visualize NBIM Wealth Dynamics
Developing an interactive RShiny website to visualize the wealth dynamics of the Norwegian Government Pension Fund Global (NBIM) serves several important purposes.
Firstly, it provides an intuitive and accessible tool for understanding how one of the world’s largest sovereign wealth funds allocates and grows its assets over time. By visualizing investment trends, asset distribution, and historical performance, users—whether individuals, researchers, or policymakers—can gain valuable insights into long-term, sustainable investment strategies.
Secondly, the platform can serve as an educational resource, promoting awareness of responsible and diversified investment practices. NBIM’s disciplined approach to balancing risk and return, while maintaining a focus on long-term value and ethical standards, offers a practical model for both institutional and personal investment planning.
Ultimately, the website supports better financial decision-making and fosters a deeper appreciation for sustainable investment philosophy in a global context.
7.2 Features Overview
7.2.1 Dashboard Structure
The dashboard consists of five main sections: Value Development, Equity, Fixed Income, Real Estate, and Infrastructure.
Value Development serves as an overview of the NBIM. This section introduces the background and purpose of the fund, and provides a comprehensive visualization of its overall asset dynamics. It includes interactive plots that illustrate:
- Value Growth over time
- Asset Holdings across different classes
- Annual Returns of the fund
These visualizations offer users a macro-level understanding of how the fund operates, how it has evolved financially, and how its investment strategies have performed across time. This section is particularly useful for gaining an initial and holistic perspective before diving into asset-specific analyses.
7.3 Value Development
7.3.1 Map Controls
The dashboard features an interactive world map that allows users to explore NBIM’s global investment distribution across different years.
7.3.1.1 Year Selector: View investment distribution for a specific year.
7.3.1.2 Currency selector:
7.3.1.3 Click on the Map
By clicking on themap, the website displays the four types of investment of NBIM in the selected country.
7.3.2 Value Growth
All plots in the Rshiny project are generated by
plotlyorggplotly, which means that the user owns a variety of options to interact with all plots.
7.3.2.1 Intro
The section firstly proveides an introduction of four value types.
7.3.2.2 Year selection
User can select the start year of charts below.
7.3.2.3 Development of the Norwegian Government Pension Fund Global (NBIM)
- The plot shows the stacked components of the fund’s growth: Krone rate, Return, and Inflows (after management costs), with the Total represented by a gray line.
- The right plot demonstrates the interactive functionality of the dashboard. When hovering over a specific year (e.g., 2023), detailed data such as total value and component breakdowns are displayed dynamically as tooltips.
7.3.2.4 Value Development by asset class
1. Value Development
2. Value Breakdown
3. Percentage of Market Value
7.3.3 Holdings
Similarily, all of the plots are generated by
plotlyorggplotly, so they are interactive and the interations are demonstrated as before.
7.3.4 Returns
- Start Year Selection
- Annual Return Plot
7.3.5 Relative Returns
7.4 Equity
Equity represents ownership in a company, typically in the form of stocks or shares. When an investor purchases equity in a company, they acquire a claim on part of the company’s assets and earnings. Equity investments are generally considered higher risk than fixed-income investments but also offer the potential for higher returns.
7.4.1 Explanation of the map
The control panels are set to be transparent, and turn white when the mouse hovers over them, providing a clear visual cue to the user that the panel is active or selectable.
The control panel on the left side is used to select the year for the map visualization. It also displays key information corresponding to the selected year, including:
- The return rate of the fund in that year
- The proportion of investments allocated to each continent
- The distribution of investments across different regions within that year
This layout allows users to explore not only the geographic spread of investments but also the performance and strategic focus of the fund over time.
- When the mouse hovers over a specific location on the map, the equity holdings for that region are displayed. The displayed values update dynamically according to the selected currency.
- The floating panel in the bottom-right corner allows users to select the currency type. It also provides a legend that explains how the color intensity on the map corresponds to the amount of investment.
7.5 Fixed Income
Fixed income refers to investment instruments that provide regular, predictable returns in the form of interest or dividend payments. These assets typically include government bonds, corporate bonds, and other debt securities. Unlike equities, fixed income investments do not represent ownership in a company but rather a loan made by the investor to a borrower.
The map for Fixed Income functions similarly to that of Equity in most aspects, so I will only introduce the differences.
This chart shows the Fixed Income holdings by instrument. The chart breaks down the fund’s fixed income investments based on the type of financial instruments they are allocated to.
From the chart, we can see that the proportions of the three types are nearly identical, indicating that NBIM adopts a relatively simple strategy of broadly distributing its fixed income investments.
7.6 Real Estate
Real estate refers to investments in physical properties such as office buildings, shopping centers, logistics facilities, and other commercial or residential real estate assets. In the context of large institutional investors like NBIM, real estate investments are typically made through direct ownership or partnership structures in unlisted (private) markets.
NBIM began including real estate in its portfolio in 2011, focusing on high-quality assets in major global cities. This asset class serves as a hedge against inflation and adds a layer of resilience to the fund’s overall risk-return profile.
7.7 Infrastructure
Infrastructure refers to investments in large-scale physical systems and facilities that are essential for the functioning of an economy. For NBIM, infrastructure—especially in renewable energy—has become a growing area of focus in recent years. These investments support both financial returns and the fund’s commitment to sustainability and responsible investment.
